python - 去除验证码干扰线用什么算法好

Question

去除诸如以上验证码干扰线，采用 灰度图 → rof算法 → ImageEnhance.Sharpness → 二值图 产生如下效果

部分孤立噪点难以去除干净，这也对tesseract的识别产生影响，上面的方式只有30%识别率。

Answer 1

For this kind of verification code:
1. The interference line is thin, only about 1px-2px
2. The font is thicker than the interference line

You can filter points like this after binarization:
1. There are less than 2 1 points among the 8 surrounding points